Closed gclsoft closed 4 years ago
我的FaceRecognizer 复制了https://github.com/StevenLei2017/AIprojects/blob/7cabc6f3773e87f46d54c6d5afa37313de075eb1/007%E4%BA%BA%E8%84%B8%E8%AF%86%E5%88%AB%E5%AE%9E%E8%B7%B5_mxnet%E7%89%88insightFace/code/FaceRecognizer_2.py
# 加载常用的库 import numpy as np import os import cv2 from PIL import Image import time # 加载mxnet库 import mxnet as mx import mxnet.ndarray as nd # 加载用于标准化的sklearn.preprocessing from sklearn import preprocessing from sklearn.model_selection import KFold # 加载人脸向量化模型 def load_model(prefix='C:\\ai\\new\\insighrface_camera_gcl\\test_run\\resources\\insightFace_model\\model', epoch=0, batch_size=10): symbol, arg_params, auxiliary_params = mx.model.load_checkpoint(prefix, epoch) all_layers = symbol.get_internals() output_layer = all_layers['fc1_output'] context = mx.cpu()#gpu(0) model = mx.mod.Module(symbol=output_layer, context=context, label_names=None) model.bind(data_shapes=[('data', (batch_size, 3, 112, 112))]) model.set_params(arg_params, auxiliary_params) return model # 定义类 人脸向量化器FaceVectorizer class FaceVectorizer(object): # 实例化对象后的初始化方法 def __init__(self): self.batch_size = 1 self.model = None self.database_2d_array= None # 获取输入数据 def get_feedData(self, image_4d_array): image_list = [] for image_3d_array in image_4d_array: height, width, _ = image_3d_array.shape if height != 112 or width != 112: image_3d_array = cv2.resize(image_3d_array, (112, 112)) image_list.append(image_3d_array) image_4d_array_1 = np.array(image_list) image_4d_array_2 = np.transpose(image_4d_array_1, [0, 3, 1, 2]) image_4D_Array = nd.array(image_4d_array_2) image_quantity = len(image_list) label_1D_Array = nd.ones((image_quantity,)) feed_data = mx.io.DataBatch(data=(image_4D_Array,), label=(label_1D_Array,)) return feed_data # 使用insightFace模型，把多张人脸的图像数据转换多张人脸的特征向量 # 返回结果为2维数组 def get_feature_2d_array(self, image_4d_array): if len(image_4d_array.shape) == 3: image_4d_array = np.expand_dims(image_4d_array, 0) assert len(image_4d_array.shape) == 4, 'image_ndarray shape length is not 4' image_quantity = len(image_4d_array) if image_quantity != self.batch_size or not self.model: self.batch_size = image_quantity self.model = load_model(batch_size=self.batch_size) feed_data = self.get_feedData(image_4d_array) self.model.forward(feed_data, is_train=False) outputs = self.model.get_outputs() output_2D_Array = outputs[0] output_2d_array = output_2D_Array.asnumpy() feature_2d_array = preprocessing.normalize(output_2d_array) return feature_2d_array # 计算准确率 def get_accuracy(distance_1d_array, actual_isSame_1d_array, threshold): predict_isSame_1d_ndarray = np.less(distance_1d_array, threshold) true_positive_quantity = np.sum(np.logical_and( predict_isSame_1d_ndarray, actual_isSame_1d_array)) true_negetive_quantity = np.sum(np.logical_and( np.logical_not(predict_isSame_1d_ndarray), np.logical_not(actual_isSame_1d_array))) accuracy = float(true_positive_quantity + true_negetive_quantity) / len(distance_1d_array) return accuracy # 获取2个特征2维数组间的距离 import math def get_distance_1d_array(feature_2d_array_1, feature_2d_array_2, distance_method='euclidean'): # 计算欧几里得距离 if distance_method == 'euclidean': diffValue_2d_array = np.subtract(feature_2d_array_1, feature_2d_array_2) distance_1d_array = np.sum(np.square(diffValue_2d_array), 1) # 计算余弦相似度 elif distance_method == 'cosine': numerator_1d_array = np.sum(np.multiply(feature_2d_array_1, feature_2d_array_2), axis=1) denominator_1d_array = np.linalg.norm(feature_2d_array_1, axis=1) * np.linalg.norm(feature_2d_array_2, axis=1) similarity_1d_array = numerator_1d_array / denominator_1d_array distance_1d_array = np.arccos(similarity_1d_array) / math.pi return distance_1d_array # 定义类 人脸识别器FaceRecognizer class FaceRecognizer(object): # 实例化对象后的初始化方法 def __init__(self, face_dirPath='../resources/face_database', distance_method='euclidean'): self.feature_dimension = 512 self.distance_method = distance_method self.face_vectorizer = FaceVectorizer() self.fileSuffix_set = set(['jpg', 'bmp', 'png']) self.load_database(face_dirPath) # 加载人脸数据库 def load_database(self, face_dirPath='./face_database'): # 统计人脸数据库的人数，人脸图像数量 try: self.personName_list = next(os.walk(face_dirPath))[1] personId_list = [] for i, personName in enumerate(self.personName_list): dirPath = os.path.join(face_dirPath, personName) fileName_list = next(os.walk(dirPath))[2] for fileName in fileName_list: fileSuffix = os.path.splitext(fileName)[1][1:] if fileSuffix in self.fileSuffix_set: personId_list.append(i) self.personId_1d_array = np.array(personId_list) self.bincount_1d_array = np.bincount(self.personId_1d_array) self.person_quantity = len(self.personName_list) self.image_quantity = len(personId_list) print('人脸数据库中总共有%d个人, %d个人脸图像' % (self.person_quantity, self.image_quantity)) # 加载人脸图像数据，转换为向量 startTime = time.time() batch_size = 30 imageData_list = [] count = 0 self.database_2d_array = np.empty((self.image_quantity, self.feature_dimension)) # 遍历每个人的人脸图像文件夹 for personName in self.personName_list: dirPath = os.path.join(face_dirPath, personName) fileName_list = next(os.walk(dirPath))[2] for fileName in fileName_list: fileSuffix = os.path.splitext(fileName)[1][1:] # 文件名后缀需要符合要求 if fileSuffix in self.fileSuffix_set: filePath = os.path.join(dirPath, fileName) image_3d_array = np.array(Image.open(filePath)) image_3d_array = cv2.resize(image_3d_array, (112, 112)) imageData_list.append(image_3d_array) count += 1 if count % batch_size == 0: image_4d_array = np.array(imageData_list) self.database_2d_array[count - batch_size: count] = self.face_vectorizer.get_feature_2d_array( image_4d_array) imageData_list.clear() if count % batch_size != 0: image_4d_array = np.array(imageData_list) remainder = count % batch_size self.database_2d_array[count - remainder: count] = self.face_vectorizer.get_feature_2d_array(image_4d_array) # 打印加载人脸图像数据花费的时间 usedTime = time.time() - startTime print('加载%d张人脸图像，总共用时 %.4f秒' % (self.image_quantity, usedTime)) # 计算得出最佳阈值 self.make_bestThreshold() except Exception as e: print('load_database error:', e) # 通过生成随机的数据集，使用10折交叉验证得出最佳阈值 def make_bestThreshold(self): self.make_randomDataSet() startTime = time.time() k_fold = KFold(n_splits=10, shuffle=False) sample_quantity = len(self.distance_1d_array) index_1d_array = np.arange(sample_quantity) # 在200个阈值中找出最佳阈值 bestThreshold_list = [] for fold_index, (train_1d_array, test_1d_array) in enumerate(k_fold.split(index_1d_array)): train_distance_1d_array = self.distance_1d_array[train_1d_array] train_isSame_1d_array = self.isSame_1d_array[train_1d_array] test_distance_1d_array = self.distance_1d_array[test_1d_array] test_isSame_1d_array = self.isSame_1d_array[test_1d_array] accuracy_list = [] threshold_1d_array = np.arange(0, 2.0, 0.01) for threshold in threshold_1d_array: train_accuracy = get_accuracy(train_distance_1d_array, train_isSame_1d_array, threshold) test_accuracy = get_accuracy(test_distance_1d_array, test_isSame_1d_array, threshold) # 训练集权重0.4，测试集权重0.6 accuracy = 0.4 * train_accuracy + 0.6 * test_accuracy accuracy_list.append(accuracy) bestThreshold_index = np.argmax(accuracy_list) bestThreshold = threshold_1d_array[bestThreshold_index] max_accuracy = np.max(accuracy_list) print('第%d次计算，使用判断阈值%.2f获得最大准确率%.4f' % (fold_index + 1, bestThreshold, max_accuracy)) bestThreshold_list.append(bestThreshold) self.bestThreshold = np.mean(bestThreshold_list) print('经过10折交叉验证，获得最佳判断阈值%.4f' % self.bestThreshold) usedTime = time.time() - startTime print('获取最佳判断阈值，用时%.2f秒' % (usedTime)) # 生成随机的数据集 def make_randomDataSet(self): startTime = time.time() sample_quantity = 32 * int(self.image_quantity ** 0.58) feature_2d_array_1 = np.empty((sample_quantity, self.feature_dimension)) feature_2d_array_2 = np.empty((sample_quantity, self.feature_dimension)) self.isSame_1d_array = np.empty((sample_quantity)) # 数据集前半部分为相同人脸 same_sample_quantity = int(sample_quantity / 2) selected_personId_1d_array = np.where(self.bincount_1d_array >= 2)[0] same_personId_1d_array = np.random.choice(selected_personId_1d_array, same_sample_quantity) for i, personId in enumerate(same_personId_1d_array): selected_index = np.where(self.personId_1d_array == personId)[0] index_1, index_2 = np.random.choice(selected_index, 2, replace=False) feature_2d_array_1[i] = self.database_2d_array[index_1] feature_2d_array_2[i] = self.database_2d_array[index_2] self.isSame_1d_array[:same_sample_quantity] = True # 数据集后半部分为不同人脸 difference_sample_quantity = int(sample_quantity / 2) index_1d_array = np.arange(self.image_quantity) for i in range(same_sample_quantity, same_sample_quantity + difference_sample_quantity): index_1, index_2 = np.random.choice(index_1d_array, 2) personId_1, personId_2 = self.personId_1d_array[[index_1, index_2]] # 通过循环判断，使personId_1不等于personId_2 while personId_1 == personId_2: index_1, index_2 = np.random.choice(index_1d_array, 2) personId_1, personId_2 = self.personId_1d_array[[index_1, index_2]] feature_2d_array_1[i] = self.database_2d_array[index_1] feature_2d_array_2[i] = self.database_2d_array[index_2] self.isSame_1d_array[same_sample_quantity:] = False # 打印随机生成数据集花费的时间 usedTime = time.time() - startTime print('随机生成数据集，总共%d组人脸，用时%.2f秒' % (sample_quantity, usedTime)) # 得出2个特征2维数组间的距离 self.distance_1d_array = get_distance_1d_array(feature_2d_array_1, feature_2d_array_2, self.distance_method) # 获取人脸对应的人名 def get_personName(self, imageFilePath): image_3d_array = np.array(Image.open(imageFilePath)) personName = self.get_personName_1(image_3d_array) return personName # 获取人脸对应的人名 def get_personName_1(self, image_3d_array): try: if len(image_3d_array.shape) == 3: image_4d_array = np.expand_dims(image_3d_array, 0) elif len(image_3d_array.shape) == 4: image_4d_array = image_3d_array else: raise ValueError('传入图像数据的维度既不是3维也不是4维') feature_2d_array = self.face_vectorizer.get_feature_2d_array(image_4d_array) if not self.database_2d_array: return '找不到' distance_1d_array = get_distance_1d_array(self.database_2d_array, feature_2d_array, self.distance_method) isSame_1d_array = np.less(distance_1d_array, self.bestThreshold) predict_personId_1d_array = self.personId_1d_array[isSame_1d_array] if len(predict_personId_1d_array) == 0: personName = '无效人脸' else: min_distance_index = np.argmin(distance_1d_array) similar_personId = self.personId_1d_array[min_distance_index] predict_bincount_1d_array = np.bincount(predict_personId_1d_array) similar_percent = predict_bincount_1d_array[similar_personId] / self.bincount_1d_array[similar_personId] if similar_percent >= 0.7: personName = self.personName_list[similar_personId] else: personName = '无效人脸' return personName except Exception as ex: print(ex) return 'Unkown'
我的FaceRecognizer 复制了https://github.com/StevenLei2017/AIprojects/blob/7cabc6f3773e87f46d54c6d5afa37313de075eb1/007%E4%BA%BA%E8%84%B8%E8%AF%86%E5%88%AB%E5%AE%9E%E8%B7%B5_mxnet%E7%89%88insightFace/code/FaceRecognizer_2.py